Text Mining Technologies for Database Curation

Fabio Rinaldi

Institute of Computational Linguistics, University of Zurich, Zurich, Switzerland

Keywords:

Biomedical Text Mining, Information Extraction, Literature-based Discovery.

Abstract:

Text mining technologies, coupled with advanced user interfaces, have a great potential in the life sciences,

for example supporting the process of database curation. We present a system which has achieved competitive

results in several community-organized evaluations of text mining technologies and we discuss how such

technologies can be integrated in a curation workﬂow.

1 INTRODUCTION

Biomedical text mining is a discipline that has been

developing quite extensively in recent years. Its aim

is to automatically analyze biomedical text, and in

particular the scientiﬁc literature. Several techniques

from the ﬁelds of Computational Linguistics, Natu-

ral Language Processing (NLP) and Information Re-

trieval have been adopted for this purpose. Some

of the software tools developed by researchers in

biomedical text mining have the potential to support

the process of database curation, but can of course be

used also for other purposes, such as to enhance the

access to the information contained in the biomedical

literature for different target groups, not only biologi-

cal researchers, but also the general public. Biomedi-

cal text mining is also of great relevance for the phar-

maceutical industry. On average, it costs about 1

billion dollars to develop a completely new medici-

nal drug, and it involves the work of hundreds of re-

searchers, collectively investing in average more than

7 million hours of work on thousands of experiments.

Automated detection of previous information from

the literature can help better target such experiments,

for example by pointing to similar experiments con-

ducted in the past, or suggesting novel experiments on

the basis of previous results. This in turn is going to

have a major economical beneﬁt, freeing up resources

for novel research.

Text mining technologies are increasingly pro-

viding an effective response to the growing demand

for faster access to the vast amounts of informa-

tion hidden in the literature. Several tools are be-

coming available which offer the capability to mine

the literature for speciﬁc information, such as for

example protein-protein interactions or drug-disease

relationships. Examples of well known biomedi-

cal text mining tools are MetaMap (Aronson and

Lang, 2010), MedEvi (Kim et al., 2008), WhatIzIt

(Rebholz-Schuhmann et al., 2008), Gimli (Campos

et al., 2013), iHOP (Hoffmann and Valencia, 2004;

Hoffmann, 2007), cTAKES (Savova et al., 2010),

Open Biomedical Annotator (Jonquet et al., 2009).

The biomedical text mining community regularly

veriﬁes the progress of the ﬁeld through competi-

tive evaluations, such as BioCreative (Arighi et al.,

2011; Krallinger et al., 2011), BioNLP (Kim et al.,

2011; Cohen et al., 2009), i2b2 (Sun et al., 2013),

CALBC (Rebholz-Schuhmann et al., 2011), CLEF-

ER (Rebholz-Schuhmann et al., 2013), DDI (Segura-

Bedmar et al., 2011), BioASQ (Androutsopoulos,

2013), etc. Each of these competitions targets dif-

ferent aspects of the problem, such as detection of

mentions of speciﬁc entities (e.g. genes and chem-

icals), detection of protein interactions, assignment

of Gene Ontology tags (BioCreative), detection of

structured events (BioNLP), information extraction

from clinical text (i2b2), large-scale entity detection

(CALBC), multilingual entity detection (CLEF-ER),

drug-drug interactions (DDI), question answering in

biology (BioASQ).

The OntoGene group

at the Institute of Computa-

tional Linguistics of the University of Zurich has de-

veloped a platform for advanced text mining applica-

tions. The OntoGene system specializes in the detec-

tion of entities and relationships from selected cate-

gories, such as proteins, genes, drugs, diseases, chem-

icals. OntoGene sources its lexical resources from life

www.ontogene.org

544

Rinaldi F..

Text Mining Technologies for Database Curation.

DOI: 10.5220/0005174905440548

In Proceedings of the International Conference on Knowledge Discovery and Information Retrieval (SSTM-2014), pages 544-548

ISBN: 978-989-758-048-2

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

sciences databases, thus allowing a deeper connection

between the unstructured information contained in the

literature and the structured information contained in

databases. The quality of the system has been tested

several times through participation in some of the

community-organized evaluation campaigns, where it

often obtained top-ranked results.

Within the scope of the SNF-funded SASEBio

project (Semi-Automated Semantic Enrichment of the

Biomedical Literature, SNF grant 105315 130558/1,

2010-2014) we developed tools aimed at supporting

the process of database curation from the biomedical

literature, and promote a move towards assisted cu-

ration. By assisted curation we mean a combination

of text mining approaches and the work of an expert

curator, aimed at leveraging the power of text mining

systems, while retaining the high quality associated

with human expertise.

In the rest of this paper, we brieﬂy describe in

Section 2 the overall architecture of our text mining

system and mention evaluation results in the context

of community-organized shared tasks, then we illus-

trate in Section 3 the usage of the text mining system

within our integrated curation environment, providing

a discussion on assisted curation and results of collab-

orations with major life science databases.

2 THE OntoGene TEXT MINING

SYSTEM

In this section we provide a brief description of the

OntoGene text mining environment, with a speciﬁc

focus on its application to the detection of interac-

tions. The ﬁrst step in order to process a collec-

tion of biomedical literature consists in the annota-

tion of names of relevant domain entities in biomed-

ical literature (we consider in particular proteins,

genes, species, experimental methods, and cell lines)

and grounding them to widely accepted identiﬁers

(IDs) assigned by reference knowledge bases, such as

UniProt, EntrezGene, Cell Line Knowledge Base, etc.

The term annotation uses a large term list that is com-

piled on the basis of the entity names extracted from

the knowledge bases. This resulting list covers the

common expression of the terms. A term normaliza-

tion step is used to match the terms with their actual

representation in the text, taking into account a num-

ber of possible surface variations. Finally, a disam-

biguation step resolves the ambiguity (i.e. multiple

IDs proposed by the annotator) of the matched terms

(Rinaldi et al., 2011).

In order to account for possible surface variants

between the terms in the term list and the token se-

quences in the text, a normalization step is included in

the annotation procedure. The same normalization is

applied to the known terms of the term list and to the

candidate terms in the input text, so that a matching

between variants becomes possible despite the differ-

ences in the surface strings. In case the normalized

strings match exactly, the input sequence is annotated

with the IDs of the matching term.

Using the information concerning mentions of rel-

evant domain entities, derived as described above,

and their corresponding unique identiﬁers obtained by

the process of disambiguation, it is possible to cre-

ate candidate interactions using the co-occurrence of

two entities in a given text span (typically a sentence,

or observation window). However, using simple co-

occurrence leads to low-precision extraction of inter-

actions. In order to obtain better precision it is neces-

sary to take into account the syntactic structure of the

sentence, and other structural information. We parse

relevant sentences (containing at least two entities)

with a dependency parser, which has been adapted

to and evaluated on the biomedical domain (Schnei-

der et al., 2007). After parsing, we collect all syn-

tactic connections that exist between all the terms as

follows. For each term-coocurrence a collector tra-

verses the tree from one term up to the lowest com-

mon mother node, and down the second term, record-

ing all intervening nodes. Paths which are extracted

from the corpus can directly be used for interaction

detection. The ranking of relation candidates can be

further optimized if we apply a supervised machine

learning method. First we automatically identify the

noisy concepts that our term recognizer generates in

order to penalize them. Second, we need to adapt to

highly-ranked false positive relations which are gen-

erated by our frequency based approach. The goal is

to identify some global preference oder biases which

can be found in the reference database.

We have been active in the area of biomedical text

mining since 2005, participating in several compet-

itive evaluations, and often obtaining top-ranked re-

sults. Some signiﬁcant examples are the following:

best results in ﬁnding mentions of experimental meth-

ods in BioCreative 2006 (Rinaldi et al., 2008), best re-

sults in detecting protein-protein interactions from the

literature in BioCreative 2009 (Rinaldi et al., 2010),

best results in detecting some entity types (genes and

diseases in particular) in the CALBC competition

(Rebholz-Schuhmann et al., 2010), best overall re-

sults in the triage task of BioCreative 2012 (Rinaldi

et al., 2013a). In addition we have been actively pro-

moting the idea that advanced text mining technolo-

gies can be a helpful support tool within the context of

a curation workﬂow, as described in the next section.

TextMiningTechnologiesforDatabaseCuration

545

Figure 1: ODIN screenshot. Example of visualization of text mining results using the ODIN interface. The panel on the left

shows the document with annotations, the panel on the right the corresponding concepts. The two panels are interconnected

by the interface logic: whenever an item is selected in the concept panel, the corresponding terms are highlighted in the

document panel.

3 ASSISTED CURATION

The main motivation behind biomedical curation ac-

tivities is “to help the life sciences community make

sense of all the data that is accumulating” (Bairoch,

2009). Although human curation offers the best guar-

antee of high quality results, it suffers from severe

bottlenecks which have long been recognized in the

curation community. The most pressing problem is

that of efﬁciency of the process: despite the fact that

typically several databases attempt to focus on a par-

ticular type of biological data, and often collaborate at

least sufﬁciently to prevent duplication of effort and

ensure compatibility of resulting data formats, it is

impossible for human curators to keep up with the

growing pace of publication: “Nobody will ever be

able to manually annotate all the macromolecular bi-

ological entities that exist on this planet, and conse-

quently automatization is the only solution” (Bairoch,

2009).

However, automated text mining tools cannot of-

fer sufﬁcient reliability to be applied indiscriminately

without human supervision. Therefore, the ideal solu-

tion is to combine the best capabilities of automated

systems with human supervision by highly qualiﬁed

domain experts. For this type of application, it is

necessary to develop user friendly interfaces that will

make text mining tools directly usable by curators,

rather than hinder their work with technical com-

plexities and poorly presented results. The Onto-

Gene group has implemented a user-friendly curation

framework called ODIN (Ontogene Document IN-

spector, see ﬁgure 1), which aims at satisfying several

of these requirements. Since every curation group has

speciﬁc interests and needs, it cannot be expected that

generic text mining solutions will be able to provide a

satisfactory solution to all of them. Instead, we intend

to provide a generic framework that can be then cus-

tomized to the speciﬁc requirements of each group.

The usage of ODIN as a curation tool has been

tested in a few collaborations with curation groups,

including PharmGKB (Rinaldi et al., 2012a), CTD

(Rinaldi et al., 2012b), RegulonDB (Gama-Castro

et al., 2013; Rinaldi et al., 2013b; Gama-Castro et al.,

2014), which are brieﬂy described here.

PharmGKB (The Pharmacogenomics Knowledge

Base) (Sangkuhl et al., 2008) is a NIH-funded, pub-

licly available online resource, developed and main-

tained at Stanford University, which aims at curating

information pertaining to the effect of genetic vari-

ants in susceptibility to diseases and drugs. They cu-

rate publications which are related to genetic variants,

and they annotate interactions between genes, drugs

and diseases. In 2011 we performed an experiment

in assisted curation in collaboration with PharmGKB,

which demonstrated the high usability level of ODIN

KDIR2014-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval

546

for a real-world curation task, as well as proved the

quality of the results of the underlying text mining al-

gorithms (Rinaldi et al., 2012a).

In 2012, as part of our participation in the BioCre-

ative 2012 text mining evaluation campaign, we an-

alyzed data provided by the Comparative Toxicoge-

nomics Database (CTD) (Davis et al., 2011), which

aims at collecting information related to the health ef-

fects of environmental chemicals. They curate rela-

tionships among chemicals, genes and diseases. We

adapted our text mining pipeline to their speciﬁc pur-

poses, and created for them a customized version of

our ODIN tool. Once again, we obtained the best re-

sults in the competition, as described in detail in (Ri-

naldi et al., 2013a).

RegulonDB (Gama-Castro et al., 2011) is another

major biological database, focusing on regulatory in-

teractions of one model organism (E. coli). Regu-

lonDB is the primary reference database of the best-

known regulatory network of any free-living organ-

ism. A collaboration between OntoGene and Regu-

lonDB was initiated in 2013 and resulted in a joint

participation in the BioCreative 2013 interactive cura-

tion task. Once again, ODIN was customized for the

speciﬁc needs of the RegulonDB database and was

tested by RegulonDB curators. Novel sentence ﬁl-

ters were implemented, which allow the curators to

see only the sentences which satisfy a given logical

condition. The joint experiment described in (Gama-

Castro et al., 2014) showed that the usage of such ﬁl-

ters allows an expert curator to reduce the amount of

text material under consideration by a factor of 10,

without any loss of accuracy in the curated results.

4 CONCLUSION

We have presented an advanced text mining architec-

ture (OntoGene Text Miner), which is embedded in

a user-friendly curation interface (ODIN). The Onto-

Gene Text Miner has been evaluated in a number of

competitive evaluation tasks and shown to perform at

state-of-the-art levels.

Besides, it has already been tested in a number of

real-world curation tasks in collaboration with major

life sciences databases.

ACKNOWLEDGEMENTS

The OntoGene group is partially supported by the

Swiss National Science Foundation (grants 100014 −

118396/1 and 105315 − 130558/1) and by Hoffman-

La Roche Pharmaceuticals, Basel, Switzerland.

REFERENCES

Androutsopoulos, I. (2013). A challenge on large-scale

biomedical semantic indexing and question answer-

ing. In BioNLP workshop (part of the ACL Confer-

ence).

Arighi, C., Roberts, P., Agarwal, S., Bhattacharya, S.,

Cesareni, G., Chatr-aryamontri, A., Clematide, S.,

Gaudet, P., Giglio, M., Harrow, I., Huala, E.,

Krallinger, M., Leser, U., Li, D., Liu, F., Lu, Z., Mal-

tais, L., Okazaki, N., Perfetto, L., Rinaldi, F., Sae-

tre, R., Salgado, D., Srinivasan, P., Thomas, P., Toldo,

L., Hirschman, L., and Wu, C. (2011). Biocreative iii

interactive task: an overview. BMC Bioinformatics,

12(Suppl 8):S4.

Aronson, A. R. and Lang, F. M. (2010). An overview of

MetaMap: historical perspective and recent advances.

J Am Med Inform Assoc, 17(3):229–236.

Bairoch, A. (2009). The future of annotation/biocuration.

Nature Precedings.

Campos, D., Matos, S., and Oliveira, J. L. (2013). Gimli:

open source and high-performance biomedical name

recognition. BMC Bioinformatics, 14:54.

Cohen, K. B., Demner-Fushman, D., Ananiadou, S., Pes-

tian, J., Tsujii, J., and Webber, B., editors (2009). Pro-

ceedings of the BioNLP 2009 Workshop. Association

for Computational Linguistics, Boulder, Colorado.

Davis, A., King, B., Mockus, S., Murphy, C., Saraceni-

Richards, C., Rosenstein, M., Wiegers, T., and

Mattingly, C. (2011). The comparative toxicoge-

nomics database: update 2011. Nucleic Acids Res.,

39(Database issue):D1067–72.

Gama-Castro, S., Rinaldi, F., Lpez-Fuentes, A., Balderas-

Martnez, Y. I., Clematide, S., Ellendorff, T. R., and

Collado-Vides, J. (2013). Assisted curation of growth

conditions that affect gene expression in e. coli k-12.

In Proceedings of the Fourth BioCreative Challenge

Evaluation Workshop, volume 1, pages 214–218.

Gama-Castro, S., Rinaldi, F., Lpez-Fuentes, A., Balderas-

Martnez, Y. I., Clematide, S., Ellendorff, T. R.,

Santos-Zavaleta, A., Marques-Madeira, H., and

Collado-Vides, J. (2014). Assisted curation of regu-

latory interactions and growth conditions of OxyR in

E. coli K-12. Database: The Journal of Biological

Databases and Curation, bau049.

Gama-Castro, S., Salgado, H., Peralta-Gil, M., Santos-

Zavaleta, A., Muniz-Rascado, L., Solano-Lira, H.,

Jimenez-Jacinto, V., Weiss, V., Garcia-Sotelo, J. S.,

Lopez-Fuentes, A., Porron-Sotelo, L., Alquicira-

Hernandez, S., Medina-Rivera, A., Martinez-Flores,

I., Alquicira-Hernandez, K., Martinez-Adame, R.,

Bonavides-Martinez, C., Miranda-Rios, J., Huerta,

A. M., Mendoza-Vargas, A., Collado-Torres, L.,

Taboada, B., Vega-Alvarado, L., Olvera, M., Olvera,

L., Grande, R., Morett, E., and Collado-Vides, J.

(2011). RegulonDB version 7.0: transcriptional reg-

ulation of Escherichia coli K-12 integrated within ge-

netic sensory response units (Gensor Units). Nucleic

Acids Res., 39(Database issue):98–105.

Hoffmann, R. (2007). Using the iHOP information resource

to mine the biomedical literature on genes, proteins,

TextMiningTechnologiesforDatabaseCuration

547

and chemical compounds. Curr Protoc Bioinformat-

ics, Chapter 1:Unit1.16.

Hoffmann, R. and Valencia, A. (2004). A gene network for

navigating the literature. Nature Genetics, 36:664.

Jonquet, C., Shah, N. H., and Musen, M. A. (2009). The

open biomedical annotator. Summit on Translat Bioin-

forma, 2009:56–60.

Kim, J., Pezik, P., and Rebholz-Schuhmann, D. (2008).

Medevi: Retrieving textual evidence of relations be-

tween biomedical concepts from medline. Bioinfor-

matics, 24(11):1410–1412.

Kim, J., Pyysalo, S., Ohta, T., Bossy, R., Nguyen, N., and

Tsujii, J. (2011). Overview of bionlp shared task 2011.

ACL HLT 2011, page 1.

Krallinger, M., Vazquez, M., Leitner, F., Salgado, D., Chatr-

aryamontri, A., Winter, A., Perfetto, L., Briganti, L.,

Licata, L., Iannuccelli, M., Castagnoli, L., Cesareni,

G., Tyers, M., Schneider, G., Rinaldi, F., Leaman, R.,

Gonzalez, G., Matos, S., Kim, S., Wilbur, W., Rocha,

L., Shatkay, H., Tendulkar, A., Agarwal, S., Liu, F.,

Wang, X., Rak, R., Noto, K., Elkan, C., Lu, Z., Dogan,

R., Fontaine, J.-F., Andrade-Navarro, M., and Valen-

cia, A. (2011). The protein-protein interaction tasks

of biocreative iii: classiﬁcation/ranking of articles and

linking bio-ontology concepts to full text. BMC Bioin-

formatics, 12(Suppl 8):S3.

Rebholz-Schuhmann, D., Arregui, M., Gaudan, S., Kirsch,

H., and Jimeno, A. (2008). Text processing through

Web services: calling Whatizit. Bioinformatics,

24(2):296–298.

Rebholz-Schuhmann, D., Clematide, S., Rinaldi, F.,

Kafkas, S., van Mulligen, E. M., Bui, C., Hellrich,

J., Lewin, I., Milward, D., Poprat, M., Jimeno-Yepes,

A., Hahn, U., and Kors, J. (2013). Entity recogni-

tion in parallel multi-lingual biomedical corpora: The

clef-er laboratory overview. In Forner, P., Mueller, H.,

Rosso, P., and Paredes, R., editors, Information Access

Evaluation. Multilinguality, Multimodality, and Visu-

alization, Lecture Notes in Computer Science, pages

353–367. Springer, Valencia.

Rebholz-Schuhmann, D., Jimeno, A., Li, C., Kafkas,

S., Lewin, I., Kang, N., Corbett, P., Milward, D.,

Buyko, E., Beisswanger, E., Hornbostel, K., Alex,

Kouznetsov, r., Witte, R., Laurila, J. B., Baker, C. J.,

Chen-Kuo, J., Clematide, S., Rinaldi, F., Farkas, R.,

ora, G., Hara, K., Furlong, L., Rautschka, M.,

Neves, M. L., Pascual-Montano, A., Wei, Q., Col-

lier, N., Chowdhury, F. M., Lavelli, A., Berlanga, R.,

Morante, R., Van Asch, V., Daelemans, W., Marina,

J. L., Mulligen, E. v., Kors, J., and Hahn, U. (2010).

Assessment of NER solutions against the ﬁrst and sec-

ond CALBC Silver Standard Corpus. In Semantic

Mining in Medicine, EBI, Cambridge, UK,.

Rebholz-Schuhmann, D., Yepes, A., Li, C., Kafkas, S.,

Lewin, I., Kang, N., Corbett, P., Milward, D., Buyko,

E., Beisswanger, E., Hornbostel, K., Kouznetsov, A.,

Witte, R., Laurila, J., Baker, C., Kuo, C.-J., Clematide,

S., Rinaldi, F., Farkas, R., Mora, G., Hara, K.,

Furlong, L. I., Rautschka, M., Neves, M., Pascual-

Montano, A., Wei, Q., Collier, N., Chowdhury, M.,

Lavelli, A., Berlanga, R., Morante, R., Van Asch, V.,

Daelemans, W., Marina, J., van Mulligen, E., Kors,

J., and Hahn, U. (2011). Assessment of ner solu-

tions against the ﬁrst and second calbc silver standard

corpus. Journal of Biomedical Semantics, 2(Suppl

5):S11.

Rinaldi, F., Clematide, S., Garten, Y., Whirl-Carrillo, M.,

Gong, L., Hebert, J. M., Sangkuhl, K., Thorn, C. F.,

Klein, T. E., and Altman, R. B. (2012a). Using ODIN

for a PharmGKB re-validation experiment. Database:

The Journal of Biological Databases and Curation.

Rinaldi, F., Clematide, S., and Hafner, S. (2012b). Rank-

ing of ctd articles and interactions using the onto-

gene pipeline. In Proceedings of the 2012 BioCreative

workshop, Washington D.C.

Rinaldi, F., Clematide, S., Hafner, S., Schneider, G.,

Grigonyte, G., Romacker, M., and Vachon, T. (2013a).

Using the OntoGene pipeline for the triage task

of BioCreative 2012. The Journal of Biological

Databases and Curation, Oxford Journals.

Rinaldi, F., Gama-Castro, S., Lpez-Fuentes, A., Balderas-

Martnez, Y., and Collado-Vides, J. (2013b). Digital

curation experiments for regulondb. In BioCuration

2013, April 10th, Cambridge, UK.

Rinaldi, F., Kaljurand, K., and Saetre, R. (2011). Termi-

nological resources for text mining over biomedical

scientiﬁc literature. Journal of Artiﬁcial Intelligence

in Medicine, 52(2):107–114.

Rinaldi, F., Kappeler, T., Kaljurand, K., Schneider, G.,

Klenner, M., Clematide, S., Hess, M., von Allmen,

J.-M., Parisot, P., Romacker, M., and Vachon, T.

(2008). OntoGene in BioCreative II. Genome Biol-

ogy, 9(Suppl 2):S13.

Rinaldi, F., Schneider, G., Kaljurand, K., Clematide, S., Va-

chon, T., and Romacker, M. (2010). OntoGene in

BioCreative II.5. IEEE/ACM Transactions on Com-

putational Biology and Bioinformatics, 7(3):472–480.

Sangkuhl, K., Berlin, D. S., Altman, R. B., and Klein, T. E.

(2008). PharmGKB: Understanding the effects of in-

dividual genetic variants. Drug Metabolism Reviews,

40(4):539–551. PMID: 18949600.

Savova, G. K., Masanz, J. J., Ogren, P. V., Zheng, J., Sohn,

S., Kipper-Schuler, K. C., and Chute, C. G. (2010).

Mayo clinical Text Analysis and Knowledge Extrac-

tion System (cTAKES): architecture, component eval-

uation and applications. J Am Med Inform Assoc,

17(5):507–513.

Schneider, G., Kaljurand, K., Rinaldi, F., and Kuhn, T.

(2007). Pro3Gres parser in the CoNLL domain adap-

tation shared task. In Proceedings of the CoNLL

Shared Task Session of EMNLP-CoNLL 2007, pages

1161–1165, Prague.

Segura-Bedmar, I., Martnez, P., and Snchez-Cisneros, D.

(2011). The 1st ddi extraction-2011 challenge task:

Extraction of drug-drug interactions from biomedical

texts. In Proc DDI Extraction-2011 challenge task,

pages 1–9, Huelva, Spain.

Sun, W., Rumshisky, A., and Uzuner, O. (2013). Evaluating

temporal relations in clinical text: 2012 i2b2 Chal-

lenge. J Am Med Inform Assoc, 20(5):806–813.

KDIR2014-InternationalConferenceonKnowledgeDiscoveryandInformationRetrieval

548